Key result
Multimodal pre-training with text reports slightly improves automated ECG classification over signal-only baselines.
Why the study?
Automated and remote ECG diagnosis can support timely cardiac assessments, but building accurate classification models across wide clinical scenarios remains challenging.
Does multimodal pre-training with text reports improve the performance of automated ECG deep learning classification models compared to signal-only training?
Does multimodal pre-training with text reports improve the performance of automated ECG deep learning classification models compared to signal-only training?
Absolute Event Rate: 0.755% vs 0.746%
Multimodal pre-training combining ECG signals with text reports improves the diagnostic accuracy of deep learning models for automated ECG interpretation, particularly when labeled data is scarce.
Objective: Heart diseases are the leading cause of death worldwide, and the electrocardiogram (ECG) is the primary diagnostic tool for assessing cardiac activity. Automated and remote ECG diagnosis can help the healthcare system with timely and high-quality cardiac assessments, especially for peripheral regions and rural areas. Automatic ECG classification has been extensively researched, but it is still challenging to build accurate models for such a wide spectrum of scenarios. Method: This study enhances the performance of ECG deep learning classification models using a multimodal pre-training stage with physician's reports. Results: Our approach improves the state-of-the-art model and achieves a mean F1 score of 0.755 over six categories using the full dataset, which is a relevant improvement for a relatively larger unlabeled corpus. Conclusion: The results demonstrate the potential to improve automated cardiac assessment with text pretraining.
No takes yet. Share an insight, caveat, or question.
Fernandes et al. (2024) studied ECG abnormalities (n=233,770). Multimodal pre-training with ECG signals and text reports vs. Pre-training with ECG signals only was evaluated on Mean F1 score across six ECG abnormality categories. Multimodal pre-training with text reports improved automated ECG classification, achieving a mean F1 score of 0.755 across six categories compared to 0.746 for the signal-only baseline.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: